I guess it would be easier if these three unrelated fixes would be split in three MRs as it looks like their "suspiciousness" and chances to get merged in their current form differ greatly. "ntdll: Ignore hardware breakpoint traps inside the signal stack." looks most straightforward, I guess it could just use spuriously added newline removal and better formatting of added condition to match the overall style around. "kernelbase: Handle misaligned stack in OpenThread": if doing it this way I guess implementing the whole function in asm is not needed, a wrapper which would force align the stack would do. But I am not sure whether such a workaround is a go not. We do have few cases of functions coded in asm to fix special cases, but realigning stack in one random function, while not necessarily impossible looks a bit weird. I've encountered similar cases in the past, for one, similar scenario hitting OpenMutexW() (which is currently worked around in Proton in rather ad-hoc but much shorter way). I'd think it could be ideal if we had a simple way to force such sort of stack realignment if needed, maybe generic asm wrapper macro in include/wine/asm.h if there is no better way. An attribute would be better but not sure offhand if good attribute for that exists. AFAIK there is no such to force the actual stack realignment, not sure about \__attribute_\_((optimize("no-tree-vectorize"))), don't seem like it is meant exactly for that? Even if affects movaps generation its meaning is not to prevent sse instruction but disable automatic loop vectorization? "ntdll: clobber nonvolatile-registers across VEH callbacks.": a more idiomatic way to do this would probably be making an asm wrapper (like we already have for call_seh_handler, for different reasons). -- https://gitlab.winehq.org/wine/wine/-/merge_requests/11069#note_142231